A novel site contributing to growth-arrest-specific gene 6 binding to its receptors as revealed by a human monoclonal antibody

Abstract

Gas6 (growth-arrest-specific gene 6) is a vitamin K-dependent protein known to activate the Axl family of receptor tyrosine kinases. It is an important regulator of thrombosis and many other biological functions. The C-terminus of Gas6 binds to receptors and consists of two laminin-like globular domains LG1 and LG2. It has been reported that a Ca²⁺-binding site at the junction of LG1 and LG2 domains and a hydrophobic patch at the LG2 domain are important for receptor binding [Sasaki, Knyazev, Cheburkin, Gohring, Tisi, Ullrich, Timpl and Hohenester (2002) J. Biol. Chem. 277, 44164–44170]. In the present study, we developed a neutralizing human monoclonal antibody, named CNTO300, for Gas6. The antibody was generated by immunization of human IgG-expressing transgenic mice with recombinant human Gas6 protein and the anti-Gas6 IgG sequences were rescued from an unstable hybridoma clone. Binding of Gas6 to its receptors was partially inhibited by the CNTO300 antibody in a dose-dependent manner. To characterize further the interaction between Gas6 and this antibody, the binding kinetics of CNTO300 for recombinant Gas6 were compared with independently expressed LG1 and LG2. The CNTO300 antibody showed comparable binding affinity, yet different dependence on Ca²⁺, to Gas6 and LG1. No binding to LG2 was detected. In the presence of EDTA, binding of the antibody to Gas6 was disrupted, but no significant effect of EDTA on LG1 binding was evident. Further epitope mapping identified a Gas6 peptide sequence recognized by the CNTO300 antibody. This peptide sequence was found to be located at the LG1 domain distant from the Ca²⁺-binding site and the hydrophobic patch. Co-interaction of Gas6 with its receptor and CNTO300 antibody was detected by BIAcore analysis, suggesting a second receptor-binding site on the LG1 domain. This hypothesis was further supported by direct binding of Gas6 receptors to an independently expressed LG1 domain. Our results revealed, for the first time, a second binding site for Gas6–receptor interaction.

INTRODUCTION

Gas6 (growth-arrest-specific gene 6) is a vitamin K-dependent protein that activates the Axl family of receptors, which includes Axl (Ufo/Ark), Sky (Dtk/Tyro3/Rse/Brt/Tif) and Mer (Eyk, Nyk) [1–7]. It is a 70 kDa protein with a structure similar to Protein S [8], a negative regulator of coagulation. Gas6 consists of a Gla domain, four EGF (epidermal growth factor) domains and a C-terminus consisting of two laminin-like globular domains LG1 and LG2. Gas6 shares an approx. 40% sequence similarity with Protein S. However, it lacks the thrombin cleavage site typical of vitamin K-dependent coagulation factors.

Gas6 is widely expressed in terminally differentiated cells of most organs including capillary endothelial cells, VSMCs (vascular smooth-muscle cells) and neurons [8,9]. It is also found in the alpha granules of platelets that are transported to the cell surface on activation [10,11]. Gas6 is generally not detected in the plasma, macrophages, basophils, neutrophils or peripheral lymphocytes. Under pathological conditions, Gas6 is up-regulated at sites of inflammation, vessel injury and in VSMCs of atherosclerotic plaques [12–15].

In accordance with its global distribution, Gas6 is involved in cell survival or proliferation of many cell types including endothelial cells [14], VSMCs [16,17], mesangial cells [18], osteoclasts [19], chondrocytes [20], Schwann cells [21], epithelial cells [22] and fibroblasts [23]. It is also a chemotactic factor for VSMCs [24]. Studies in Gas6 knockout mice demonstrated that Gas6 is an important platelet amplifier. Gas6-depleted platelets no longer respond to low concentrations of most agonists. As a result, Gas6 knockout mice are protected from challenges of both venous thrombosis and arterial thrombosis [11].

The ability of Gas6 to bind to and activate its receptors requires vitamin K-dependent γ-carboxylation [25,26]. However, previous studies [27,28] also indicated that truncated Gas6 or a splice variant of Gas6 containing only the C-terminal globular repeats is sufficient to activate the receptors. It was postulated that the C-terminal repeats of Gas6 are responsible for its biological activity, whereas the N-terminus modulates its activity through γ-carboxylation. Studies of the crystal structure of this C-terminus have identified a Ca²⁺-binding site and a hydrophobic patch that are important for Gas6–receptor interaction. Site-directed mutagenesis of several residues within the hydrophobic patch has been shown to diminish receptor binding [29]. It was therefore postulated that the receptor-binding site of Gas6 resides in the hydrophobic patch on the LG2 domain. In the present study, we report the identification of a second receptor-binding site on the LG1 domain as demonstrated by direct binding of the receptors to the LG1 domain. We have also identified a novel peptide sequence of Gas6 that is recognized by a neutralizing monoclonal antibody but is located on the LG1 domain outside the Ca²⁺-binding site and the hydrophobic patch.

MATERIALS AND METHODS

Generation of recombinant Gas6 containing a FLAG epitope

Human Gas6 cDNA was obtained by PCR of reverse-transcribed mRNA from the human CHRF cell line. The upstream primer used was 5′-GCTCTAGAACCATGGCCCCTTCGCTCTCGC-3′ and the downstream primer used was 5′-GCTCTAGAACAGAGACTGAGAAGCCTGC-3′. Gas6–FLAG cDNA containing a FLAG epitope at the C-terminus was then obtained by a second PCR using the same upstream primer and 5′-GCTCTAGACTACTTGTCGTCGTCGTCCTTGTAGTCGGCTGCGGCGGGCTCCACGG-3′ as the downstream primer. The full-length cDNA was then inserted into the pcDNA3.1 plasmid containing a hygromycin-resistant gene (Invitrogen Life Technologies, Carlsbad, CA, U.S.A.) to generate the plasmid vector pcDNA3.1/Gas6_Flag.

A stable cell line expressing human Gas6–FLAG was obtained by transfecting pcDNA3.1/Gas6_Flag vector into HEK-293 cells (human embryonic kidney 293 cells) followed by selection with 600 μg/ml hygromycin B. Drug-resistant colonies were picked and screened for protein expression by immunoblotting using both anti-Gas6 (Santa Cruz Biotechnology, Santa Cruz, CA, U.S.A.) and anti-FLAG antibodies (Sigma, St. Louis, MO, U.S.A.). Positive clones were maintained in the standard growth medium containing 500 μg/ml hygromycin B and 10 μg/ml vitamin K1. Conditioned media from the cell culture were stored in the presence of 5 mM benzamidine and loaded on to an anti-FLAG M2 affinity gel (Sigma). Gas6–FLAG recombinant protein was eluted from the column using the FLAG peptide according to the manufacturer's instructions. After dialysis, the protein was concentrated and subjected to analysis. The identity of purified Gas6–FLAG protein was confirmed by peptide mapping using the Voyager System 1092 (Applied Biosystems, Foster City, CA, U.S.A.), whereas the N-terminal sequence was verified using the Precise Sequencer 1092 (Applied Biosystems).

Generation of LG1 and LG2

Sequences corresponding to LG1 (317–451 amino acids) and LG2 (503–646 amino acids) were selected based on published information for the domain [8] and a simulated structure model of Gas6. DNA sequences encoding LG1 and LG2 were PCR-amplified from human Gas6 cDNA using the primers 5′-BamHI-CCAGGCTGGATGCTGAGTTTG-3′ and 5′-EcoRI-TCCGTTCAGCCAGTTCCAGCT-3′ for LG1 and primers 5′-BamHI-ATCCGCCCAGCCGCAGACA-3′ and 5′-EcoRI-CAGTGTCATGCAGCCGCGGT-3′ for LG2. The PCR products were cloned in-frame into the BamHI–EcoRI site of the pRSET bacterial expression vector (Invitrogen Life Technologies). Expression of LG1 or LG2 in the pRSET vector was under the control of an IPTG (isopropyl β-D-thiogalactoside)-inducible promoter. The N-terminus of LG1 or LG2 was ligated in-frame with a poly-histidine tag to facilitate purification. Briefly, the bacterial cell line BL21(DE3)pLysS was transformed with the expression vector and treated with 1 mM IPTG. Maximum expression was achieved at 3–4 h after IPTG induction. The cells were harvested, lysed and the protein of interest was affinity-purified according to the instruction manual for the ProBond Purification System (Invitrogen Life Technologies). Eluted proteins were desalted and the sequences were confirmed by tryptic peptide mapping using the Voyager System 1092 (Applied Biosystems).

Generation of human monoclonal antibody CNTO300

F₂ hybrid mice (CBA/J×C57/BL6/J), containing human variable and constant region antibody transgenes for both heavy and light chains [30–35], were immunized with recombinant human Gas6–FLAG protein by standard procedures. Sera from immunized animals were tested for Gas6 reactivity using ELISAs. Hybridoma cells were generated from mice producing Gas6 antibodies by standard procedures. A Gas6 reactive hybridoma clone was identified but it was unstable in cell culture. The antibody gene sequences were rescued by molecular cloning from the hybridoma cells. A GeneRacer kit (Invitrogen Life Technologies) was used to amplify the variable regions according to the manufacturer's instructions. The GeneRacer 5′-primer was used, along with the primer 5′-GTGACTTCGCAGGCGTAGACT-3′ to amplify the light chain or the primer 5′-GTACTCCTTGCCATTCAG-3′ to amplify the heavy chain. A TOPO TA cloning kit (Invitrogen Life Technologies) was used to clone both the heavy- and light-chain PCR products into the PCR 4-TOPO vector and the heavy- and light-chain sequences were identified by DNA sequencing. The heavy-chain variable region was then cloned into a human IgG₁ expression vector and the light-chain variable region was cloned into a human kappa light-chain expression vector, both driven by a cytomegalovirus promoter. The antibody, CNTO300, was expressed by transient expression of the heavy- and light-chain constructs in HEK-293E cells. The expressed antibody was purified by Protein A affinity chromatography.

BIAcore analysis

A BIAcore 2000 (Biacore, Piscataway, NJ, U.S.A.) was used to determine the binding constants of the interaction of the CNTO300 antibody with Gas6, LG1 or LG2. A capture sensor surface was prepared by covalently immobilizing rabbit anti-human IgG Fc-specific polyclonal antibody to a CM-5 chip using an NHS/EDC amine coupling kit (Biacore). Approximately 3000 resonance units of polyclonal antibody were immobilized. Binding studies were performed by equilibrating the instrument and sensor surface using a running buffer of Hepes-buffered saline containing 3 mM CaCl₂ and 3 mM MgCl₂. CNTO300 was diluted into the running buffer and then injected over the modified sensor surface. Approximately 100–150 resonance units of CNTO300 were captured. Samples of 400 nM Gas6, LG1 or LG2 were then passed over this surface at a flow rate of 30 μl/min using the inject command. To study the effect of metal on binding, the instrument was equilibrated with Hepes-buffered saline containing 3 mM EDTA. The samples were also prepared using this buffer. The data were analysed using the BIAevaluation 3.2 software using the simple 1:1 model. The data were analysed globally, using a simultaneous fit for both dissociation (k_d; s⁻¹) and association (k_a; M⁻¹·s⁻¹) and the value of K_d (M) was calculated as k_d/k_a.

BIAcore analyses of Gas6 (10 nM for Axl-Fc and Dtk-Fc and 150 nM for Mer-Fc) binding to different receptors were performed as described above, except that 50 mM of the soluble human receptors Axl-Fc, Dtk-Fc and Mer-Fc (a chimaera between the extra-cellular portion of the receptor and human IgG Fc; R & D Systems, Minneapolis, MN, U.S.A.) were used instead of the CNTO300 antibody. Similarly, binding of LG1 to Axl-Fc was measured.

To detect co-binding of Gas6 to Axl-Fc and the CNTO300 antibody, the CNTO300 antibody was immobilized on to a sensor surface using NHS/EDC coupling as described above. However, CNTO300 was diluted into 10 mM Hepes (pH 7.0) to make a 30 μg/ml solution. This solution was flowed over the sensor surface at 5 μl/min for 10 min. The sensor surface was then treated with 1 M ethanolamine (pH 8) as before. The effect of Axl-Fc on Gas6 binding with CNTO300 was shown by injecting, at 20 μl/min, a 160 μl sample solution of 200 nM Axl-Fc, 20 nM Gas6 or 20 nM of Axl-Fc and Gas6 combined. The surface was regenerated with a 15 s pulse of 50 mM NaOH.

CNTO300 binding to Gas6

A solid-phase ELISA was used to detect CNTO300 binding to recombinant human Gas6–FLAG and mouse Gas6 (R & D Systems). Briefly, a plate was coated with recombinant human or mouse Gas6 protein at 1 μg/ml overnight at 4 °C. The plate was then washed, blocked with SuperBlock (Pierce, Rockford, IL, U.S.A.) and a dose titration of CNTO300 was incubated on the plate for 2 h at room temperature. The plate was washed and incubated with a 1:2000 dilution of goat anti-human IgG Fc-specific antibody (Sigma) for 30 min to detect the bound antibody. The plate was developed using o-phenylenediamine dihydrochloride substrate and colour development was halted using equal volumes of 2 M H₂SO₄. The A₄₉₀ was measured on a multi-well spectrophotometer (Molecular Devices, Sunnyvale, CA, U.S.A.).

Neutralization of Gas6 binding to its receptor by CNTO300

A solid-phase ELISA was used to assess the ability of the CNTO300 monoclonal antibody to neutralize the binding of Gas6 to its receptor Axl. Soluble human receptor Axl-Fc (R & D Systems) was coated at 0.5 μg/ml on to a 96-well plate and incubated overnight at 4 °C. The plate was then washed, blocked with SuperBlock (Pierce) and, then, 50 ng/ml of recombinant human Gas6–FLAG in the presence of antibody was added to each well. A commercial neutralizing polyclonal antibody to human Gas6 (R & D Systems) was used as a positive control. The plate was incubated at room temperature for 2 h, and anti-FLAG M2 peroxidase conjugate (Sigma) was used as a secondary antibody to detect the bound Gas6 protein. The plate was developed using o-phenylenediamine dihydrochloride substrate and colour development was halted using equal volumes of 2 M H₂SO₄. The A₄₉₀ was measured on a multi-well spectrophotometer (Molecular Devices).

CNTO300 epitope mapping

Immunoprecipitation and tryptic digestion

A mixture of 75 μg of recombinant human Gas6–FLAG protein and 75 μg of CNTO300 monoclonal antibody was incubated at 4 °C overnight, followed by buffer exchange using a 100000 nominal molecular weight limit filter unit and a digestion buffer (0.1 M Tris/HCl buffer, pH 8.5). Proteolysis was initiated by the addition of 4 μg of trypsin and the reaction was continued for 2 h at 37 °C. After digestion, the remaining antigen–antibody complex was captured by Protein G beads, washed with PBS, and the bound peptides were eluted in 40 μl of elution buffer [0.5% trifluoroacetic acid and 10% (v/v) acetonitrile in distilled water].

MALDI–TOF (matrix-assisted laser-desorption ionization–time-of-flight) MS analysis

Eluted peptides from the trypsin-digested complex were desalted by ZipTip C18 and spotted on to a MALDI–TOF chip. MALDI–TOF (DE-STR; Applied Biosystems) was calibrated by Calibration Mixture 2 (Applied Biosystems). The mass spectra were acquired at laser intensity 1650 for the mass range 850–5000 Da. Major mass peaks were further analysed by LC tandem MS (service provided by ProtTech, Norristown, PA, U.S.A.).

LC tandem MS analysis

The digested protein sample was dried in a SpeedVac (Thermo Electron, Waltham, MA, U.S.A.) and redissolved in 0.1% trifluoroacetic acid, followed by ZipTip C18 (Millipore, Billerica, MA, U.S.A.) clean-up according to the manufacturer's instructions. The eluate from ZipTip C18 clean-up was then dried and resuspended in 0.1% trifluoroacetic acid before loading into a C18 capillary column for LC tandem MS analysis.

A Finnigan (ThermoFinnigan, San Jose, CA, U.S.A.) LCQ ion-trap mass spectrometer coupled online with a Beckman gradient HPLC system was used. A C18 capillary column (75 μm inner diameter and 10 cm length) was connected to a NanoSpray device to allow the delivery of stable electrospray in the flow rate range 100–1500 nl/min. Mobile phases were solvent A (2% acetonitrile, 97.5% water and 0.1% formic acid) and solvent B (90% acetonitrile, 9.5% water and 0.1% formic acid). The ion-trap mass spectrometer was set to operate with the automatic gain control on.

For data analysis, the tandem MS data were searched against the most recent non-redundant protein database and then manually analysed and confirmed.

RESULTS

Generation of a human anti-Gas6 monoclonal antibody

Full-length recombinant human Gas6 containing a FLAG epitope at the C-terminus was generated as the antigen to develop anti-Gas6 antibodies. The epitope-tagged Gas6 retained the bioactivity of binding to Axl, Dtk and Mer receptors (Table 1) and inducing receptor activations (results not shown). The binding constants of FLAG-tagged Gas6 are in general agreement with published data using a tag-free Gas6 [36]. Our results suggested that the addition of the FLAG epitope did not alter the ability of Gas6 to bind to the receptors and thus maintained the overall conformation of Gas6.

Table 1. Binding constants of recombinant human Gas6–FLAG to its receptors.

Using a BIAcore 2000 analyser, 50 mM of the soluble receptors Axl-Fc, Dtk-Fc or Mer-Fc were captured by immobilized rabbit anti-human IgG Fc-specific polyclonal antibody. Samples of Gas6 (10 nM for Axl and Dtk and 150 nM for Mer) were then passed over this surface and the data were analysed globally using a simultaneous fit for both dissociation (k_d) and association (k_a). The value for K_d was calculated as k_d/k_a.

	kd (s−1)	ka (M−1·s−1)	Kd (pM)
Axl-Fc	4.8×10−5	9.1×105	53
Mer-Fc	1.4×10−4	4.6×105	304
Dtk-Fc	1.7×10−5	5.4×105	31.5

Human IgG-expressing transgenic mice were used to generate Gas6 antibodies by immunization with recombinant human Gas6–FLAG protein. In general, immunization with Gas6 did not result in robust antibody titres. B-cells from an immunized transgenic mouse that demonstrated a specific human IgG titre of 1:6400 to human Gas6 were fused to murine FO myeloma cells. Only one hybridoma secreting a human antibody specific to human Gas6 was identified. From the first subcloning, only 4.5% of the clones retained Gas6 reactivity and none of the daughter clones from subcloning had Gas6 reactivity. Subsequent attempts to generate anti-Gas6 hybridoma using different immunization methods also failed. Owing to the difficulty in obtaining a stable hybridoma clone, the parental positive cells were used to rescue the antibody genes by molecular cloning. The DNA sequences of the cloned antibody variable region genes were determined. As expected, the sequences were found to be derived from the specific human variable region transgenes used in the generation of the transgenic mice. The CNTO300 antibody was expressed and purified from transfected mammalian cells.

In an ELISA, the CNTO300 antibody was capable of binding to immobilized recombinant human and mouse Gas6, with a log higher affinity to the human protein compared with the mouse protein (Figure 1). A similar result was also obtained by BIAcore binding analysis. Given the high degree of homology between the human and mouse Gas6 proteins, the cross-reactivity of CNTO300 antibody to mouse Gas6 was not surprising. The CNTO300 antibody was also found to cross-react, with a lower affinity, with human Protein S.

Figure 1. Binding of the CNTO300 antibody to recombinant human Gas6 and mouse Gas6.

In an ELISA experiment, a dose titration of CNTO300 was added to immobilized recombinant human Gas6 (■) or mouse Gas6 (○) and the binding was detected by a secondary antibody. The data points were obtained from triplicate samples and the experiments were repeated three times. O.D. 490 nm=A₄₉₀.

Neutralization of receptor binding to Gas6 by the CNTO300 antibody

The ability of the CNTO300 antibody to inhibit the binding of Gas6 to the Axl receptor was tested using ELISA. As shown in Figure 2, the CNTO300 antibody dose-dependently inhibited the binding of Gas6 to the immobilized soluble Axl receptor with an IC₅₀ value of 15 μg/ml. Similar inhibition was found using a commercial polyclonal antibody against human Gas6. No inhibition was detected using an irrelevant isotype-matched monoclonal antibody. Consistently, the soluble Axl-Fc competed for Gas6 binding to captured CNTO300 antibody by BIAcore binding analysis (results not shown).

Figure 2. Inhibition of Gas6 binding to Axl receptor by the CNTO300 antibody.

In an ELISA experiment, human Gas6–FLAG in combination with a dose titration of either the monoclonal antibody CNTO300 (○) or a commercially available polyclonal antibody (■) was added to immobilized soluble human Axl-Fc. Gas6–receptor binding was detected using an anti-FLAG antibody. The data points were obtained from triplicate samples and the experiments were repeated three times. O.D. 490 nm=A₄₉₀.

Antibody binding kinetics

BIAcore analyses were performed to determine the binding kinetics of the CNTO300 antibody to Gas6 in comparison with independently expressed LG1 and LG2 domains (Table 2 and Figure 3). The CNTO300 antibody bound to the full-length human Gas6 protein with K_d=7.68 nM. This association was disrupted by the addition of EDTA (Figure 3A), consistent with a key role for Ca²⁺ in maintaining Gas6 protein conformation. Further studies using independently expressed LG1 and LG2 showed that CNTO300 bound to the LG1 domain but not to the LG2 domain. The binding constant of the CNTO300 antibody with LG1 was comparable with the full-length Gas6 protein, yet EDTA had a minimal effect in this case (Table 2 and Figure 3B).

Table 2. Binding constants of the CNTO300 antibody for Gas6 and LG1.

Using a BIAcore 2000 analyser, the CNTO300 antibody was captured by immobilized rabbit anti-human IgG Fc-specific polyclonal antibody. Samples of 400 nM Gas6, LG1 or LG2 were then passed over this surface. To study the effect of metal on binding, the instrument was equilibrated with a buffer containing 3 mM EDTA and the samples were also prepared using this buffer. The data were analysed globally, using a simultaneous fit for both dissociation (k_d) and association (k_a). NB, no binding detected.

	kd (s−1)	ka (M−1·s−1)	Kd (nM)
Human Gas6–FLAG (−EDTA)	9.60×10−4	1.25×105	7.68
Human Gas6–FLAG (+EDTA)	NB	NB	NB
Human LG1 (−EDTA)	1.42×10−4	1.44×104	9.86
Human LG1 (+EDTA)	1.00×10−4	1.26×104	7.94
Human LG2 (−EDTA)	NB	NB	NB

Figure 3. BIAcore sensorgram demonstrating the effect of EDTA on the binding of recombinant human Gas6 and LG1 to CNTO300.

The CNTO300 antibody was captured by immobilized rabbit anti-human IgG Fc-specific polyclonal antibody. Samples of 400 nM Gas6 or LG1 were then passed over this surface. (A) Binding of Gas6 without EDTA (●) and with EDTA (△) over captured CNTO300 sensor surface. (B) Binding of Gas6 LG1 without (●) and with EDTA (△). RU, response units.

CNTO300 antibody epitope mapping

The binding epitope for CNTO300 was identified by an antibody protected protease digestion method. Briefly, human Gas6 protein was incubated with CNTO300 and then digested by trypsin. A tryptic peptide that remained bound to CNTO300 was eluted and identified by MS. The peptide showed a mass to charge ratio (m/z) of 1316.5 obtained by MALDI MS (Figure 4A). The corresponding peptide sequence, I⁴⁰³AVAGDLFQPER⁴¹⁴ located on the LG1 domain, was identified by LC tandem MS (Figure 4B). An independent mapping effort by incubation of CNTO300 with independently expressed LG1 yielded the same peptide. This result is consistent with the BIAcore binding study, demonstrating direct association of the CNTO300 antibody with the LG1 domain. Synthetic peptides containing the epitope sequence, however, bound weakly to the CNTO300 antibody and did not inhibit the binding of the CNTO300 antibody to Gas6. Therefore the epitope is not merely a linear sequence and certain structural features are required for optimal binding to the CNTO300 antibody.

Figure 4. MS analysis of tryptic Gas6 peptides bound by CNTO300.

In an antibody protected protease digestion study, recombinant human Gas6 was incubated with CNTO300 and subjected to tryptic digestion. The tryptic Gas6 peptides bound to CNTO300 were eluted and analysed by MS. (A) MALDI MS analysis of the eluted peptides. The Gas6 peptide I⁴⁰³AVAGDLFQPER⁴¹⁴ captured by the CNTO300 antibody is indicated. (B) Sequence identification of the captured Gas6 peptide, m/z=1316.5, by LC tandem MS.

The C-terminal globular repeat region of Gas6, when expressed alone, is known to be sufficient for receptor binding. The crystal structure of the C-terminus reported by Sasaki et al. [29] demonstrated a V-shaped arrangement of LG1 and LG2. This conformation seemed to be strengthened by a Ca²⁺-binding site located at the interface of the LG1 and LG2 domains. A hydrophobic patch connecting the LG1 and LG2 domains was further identified and defined as a receptor-binding site. Several mutations within the hydrophobic patch disrupted Gas6 binding to receptors. The tryptic peptide captured by the CNTO300 antibody is located at the LG1 domain but is distant from both the Ca²⁺-binding site and the putative receptor-binding hydrophobic patch (Figures 5A and 5B).

Figure 5. Putative CNTO300 antibody-binding site.

(A) Amino acid sequence of Gas6. The Gla, EGF, LG1 and LG2 domains are underlined and the expressed LG1 and LG2 fragments are shown by boldface letters. The binding site for CNTO300 is indicated by thick underlining. (B) The crystal structure of the C-terminus of Gas6 was from Sasaki et al. [29]. The Ca²⁺-binding site and the hydrophobic patch are indicated by arrows. The Gas6 tryptic peptide that bound to CNTO300 is highlighted in black.

Co-binding of Gas6 to its receptors and the CNTO300 antibody

Our previous results demonstrated that binding of the CNTO300 antibody to Gas6 partially disrupted receptor binding. Two hypotheses could be proposed. First, there might be a second receptor-binding site on the LG1 domain, and binding of receptors to this second site is blocked by the CNTO300 antibody. Secondly, binding of the CNTO300 antibody leads to conformational changes on the hydrophobic patch that could no longer support receptor binding. If the first hypothesis is correct, Gas6 should be capable of binding both the antibody and the receptor at the same time. To test this hypothesis, the ability of Gas6 to bind both the CNTO300 antibody and the receptors was tested by BIAcore analysis. The CNTO300 antibody was immobilized on the sensor surface and the ability of Gas6, Axl-Fc or Gas6+Axl-Fc to bind was measured. As expected, Gas6 bound to immobilized CNTO300 antibodies (Figure 6). Although Axl-Fc alone showed no binding, addition of Axl-Fc+Gas6 revealed co-binding of Gas6 to both the antibody and the receptors (Figure 6). This result provided strong support for the two-receptor binding-site theory.

Figure 6. BIAcore sensorgrams demonstrating the ability of a complex of Axl-Fc and Gas6 to bind to immobilized CNTO300.

△, Axl-Fc binding to the surface of immobilized CNTO300; ○, Gas6 binding to immobilized CNTO300; □, Gas6 and Axl-Fc mixed together and incubated at room temperature for 2 h. RU, response units.

Direct binding of Gas6 receptors to the LG1 domain

To confirm the presence of a second receptor-binding site on the LG1 domain, direct binding of the receptors to independently expressed LG1 domain was measured by BIAcore analysis. The Axl-Fc receptors were captured on the sensor surface and binding of Gas6, LG1 or LG2 was tested. As shown in Figure 7, the Axl-Fc receptor bound to Gas6 and, to a lesser degree, to the LG1 domain. No binding to the independently expressed LG2 domain was detected.

Figure 7. BIAcore sensorgram comparing the bindings of recombinant Gas6, LG1, and LG2 to CNTO300.

CNTO300 was captured on to the surface of immobilized rabbit anti-human Fc-specific polyclonal antibody. Samples of 50 nM each were passed over this surface; ○, Gas6; ◆, LG1; □, LG2. RU, response units.

DISCUSSION

The generation of a human monoclonal antibody to Gas6 has allowed further analysis of the Gas6–receptor interaction. The CNTO300 antibody appears to recognize a conformationally sensitive epitope. EDTA disrupted the ability of Gas6 to bind to the antibody, indicating that a Ca²⁺-sensitive conformation of Gas6 is required for antibody binding. It has been hypothesized that the Ca²⁺-binding site of Gas6 stabilizes the V-shaped conformation held by the LG1 and LG2 domains [29]. In addition, Ca²⁺ is known to influence the conformation of all γ-carboxylated proteins through interaction with the Gla domain [37]. It is probable that EDTA disrupts the conformation of Gas6 that supports the binding to CNTO300. Furthermore, synthetic peptide derived from the epitope of the CNTO300 antibody, which lacks a rigid structure, bound weakly to the antibody and failed to block binding of the antibody to Gas6.

Gas6 can only bind to its receptors if it is γ-carboxylated as a full protein, yet truncated Gas6 protein lacking the Gla domain clearly retains its biological activities [25–28]. The biological activity of Gas6, therefore, resides on the C-terminal globular repeat region, whereas the N-terminal Gla domain modulates the activity depending on the state of γ-carboxylation. In the present study, we were capable of expressing individual LG1 and LG2 domains in bacteria and showed that the LG1 domain preserved the epitope conformation for CNTO300 antibody binding. In contrast with binding to the full-length protein, EDTA did not disrupt CNTO300 binding to LG1. It is possible that the epitope conformation of Gas6, similar to that of the C-terminal region of the full protein, is modulated by the presence of LG2 in a Ca²⁺-dependent manner.

Despite the fact that the CNTO300 antibody is capable of partially blocking Gas6 binding to its receptors, both the antibody and the receptors can bind to Gas6 at the same time as demonstrated by BIAcore analysis. In the present study, we hypothesize that there are two receptor-binding sites on Gas6, the hydrophobic patch on LG2 and a novel site on LG1. The CNTO300 antibody binds to LG1 and blocks the novel receptor recognition site on the LG1 domain. This hypothesis is supported by the following data: (i) the CNTO300 antibody did not completely block receptor binding; (ii) the epitope sequence bound by the CNTO300 antibody is located on the LG1 domain distant from the Ca²⁺-binding site and the putative receptor-binding site at the hydrophobic patch on LG2; (iii) epitope mapping from the CNTO300/Gas6 or CNTO300/LG1 mixture yielded the same antibody-binding sequence, suggesting that the conformation for antibody recognition, and possibly the novel receptor-binding site, is preserved in the expressed LG1 fragment; and (iv) receptors were shown to bind to the expressed LG1 fragment directly. Binding of the CNTO300 antibody to the epitope sequence, therefore, only blocks the novel receptor-binding site on LG1, leaving the other site on LG2 still accessible for receptor binding. Further studies are required to see if the novel receptor-binding site is co-localized at the antibody-binding site.

The difficulty in generating Gas6 antibodies was unexpected. Gas6, although involved in many cellular functions, has not been directly implicated in interfering with the immune system. Yet, we were not able to generate high-titre Gas6 antibodies using the recombinant human Gas6 protein as the immunogen. The following hypotheses could be postulated. First, it is possible that Gas6 protein is not very immunogenic in rodents, especially in the human Ig transgenic mice. Further optimization of the immunization strategy may be required. Secondly, Gas6 antigen may interfere with the humoral immune response to foreign antigen. Although neither Gas6 nor its receptors were found in peripheral T- or B-cells [38], Axl and Dtk were found in myeloid leukaemic blasts and Mer was found in neoplastic T- and B-cell lines [39–41]. Studies from receptor knockout mice indicated that expression of the Axl family of receptors is required for maintenance of the immune system [42]. Mice lacking all three receptors showed enlarged spleens and developed severe autoimmunity [43]. However, a similar phenotype has not been reported in Gas6 knockout mice. It is possible that Gas6 and other yet to be identified ligand(s) work in concert to mediate the immune function through the Axl family of receptors. Thirdly, Gas6 is a known survival factor for many cell types under serum deprivation. It is possible that Gas6 is important for establishing stable hybridoma cells and anti-Gas6-expressing cells could be depleted during the hybridoma selection process. These hypotheses need to be supported by further studies directed specifically at addressing these issues.

In conclusion, we have identified a second receptor-binding site on the LG1 domain of Gas6. The interaction between Gas6 and its receptors is associated with a variety of pathological conditions in the brain, heart, kidney and other vital organs [11,13,19,44–48]. Gas6 and its receptors thus represent a new class of therapeutic targets. Understanding the nature of the Gas6–receptor interaction would ultimately help the development of novel small-molecule drugs or neutralizing monoclonal antibodies for therapeutic applications for the detrimental diseases where the interaction between Gas6 and its receptors contributes to the disease progression or pathology.